Transcript play - EnhanceEdu

Classification
Vikram Pudi
[email protected]
IIIT Hyderabad
Talk Outline

Introduction





1
Classification Problem
Applications
Metrics
Combining classifiers
Classification Techniques
The Classification Problem
Outlook
2
Temp
(F)
Humidity
(%)
Windy?
Class
play
Play Outside?
sunny
sunny
sunny
sunny
sunny
overcast
overcast
overcast
overcast
rain
rain
rain
rain
rain
75
80
85
72
69
72
83
64
81
71
65
75
68
70
70
90
85
95
70
90
78
65
75
80
70
80
80
96
true
true
false
false
false
true
false
true
false
true
true
false
false
false
sunny
77
69
true
?
rain
73
76
false
?
don’t play
don’t play
don’t play
play
Model relationship between
class labels and attributes
play
play
e.g. outlook = overcast  class = play
play
play
don’t play
don’t play
play
play
play
 Assign class labels to
new data with unknown labels
Applications

Text classification



Classify emails into spam / non-spam
Classify web-pages into yahoo-type hierarchy
NLP Problems


Risk management, Fraud detection, Computer intrusion
detection





Vision
Speech recognition
etc.
All of science & knowledge is about predicting future in terms of
past

3
Given the properties of a transaction (items purchased, amount,
location, customer profile, etc.)
Determine if it is a fraud
Machine learning / pattern recognition applications


Tagging: Classify words into verbs, nouns, etc.
So classification is a very fundamental problem with ultra-wide
scope of applications
Metrics
1.
2.
3.
4.
5.
4
accuracy
classification time per new record
training time
main memory usage (during classification)
model size
Accuracy Measure

Prediction is just like tossing a coin (random
variable X)




In statistics, a succession of independent
events like this is called a bernoulli process




5
“Head” is “success” in classification; X = 1
“tail” is “error”; X = 0
X is actually a mapping: {“success”: 1, “error” : 0}
Accuracy = P(X = 1) = p
mean value =  = E[X] = p1 + (1-p)0 = p
variance = 2 = E[(X-)2] = p (1–p)
Confidence intervals: Instead of saying
accuracy = 85%, we want to say: accuracy 
[83, 87] with a confidence of 95%
Binomial Distribution



Treat each classified record as a
bernoulli trial
If there are n records, there are n
independent and identically distributed
(iid) bernoulli trials, Xi, i = 1,…,n
Then, the random variable X = i=1,…,n Xi
is said to follow a binomial distribution


6
P(X = k) = nCk pk (1-p)n-k
Problem: Difficult to compute for large n
Normal Distribution



Continuous distribution with parameters 
(mean), 2(variance)
Probability density:
f(x) = (1/(22)) exp(-(x-)2 / (22))
Central limit theorem:



7
Under certain conditions, the distribution of the
sum of a large number of iid random variables is
approximately normal
A binomial distribution with parameters n and p is
approximately normal for large n and p not too
close to 1 or 0
The approximating normal distribution has mean μ
= np and standard deviation 2 = (n p (1 - p))
Confidence Intervals
Normal distribution with mean = 0 and variance = 1
–1



8
0
1 1.65
E.g. P[–1.65  X  1.65]
= 1 – 2  P[X  1.65] = 90%
To use this we have to
transform our random variable
to have mean = 0 and
variance = 1
Subtract mean from X and
divide by standard deviation
Pr[X  z]
z
0.1%
3.09
0.5%
2.58
1%
2.33
5%
1.65
10%
1.28
20%
0.84
40%
0.25